Statistical Analysis of Field-Based Stormwater Filtration Performance for the Ecosol Litter Basket

: Increasing speciﬁcity in water quality regulations for the discharge of stormwater to the environment has increased the requirement to more accurately characterize the performance of ﬁltration interventions. This work presents a statistical performance analysis for the Ecosol Litter Basket , an at source ﬁltration device, based on an extensive ﬁeld study. The ﬁeld evaluation of the Ecosol Litter Basket , a primary stormwater ﬁltration device, was performed over a three-year period in an urban catchment in Queensland, Australia. A total of 29 rainfall events were recorded, of which between 13 to 16 events were evaluated as qualifying for the purposes of characterizing the removal efﬁciency. A variety of pollutant removal evaluation metrics, including concentration-based and total load-based metrics, were utilized in this study to characterize the efﬁcacy of the device for removing a range of pollutants. Two approaches are proposed to facilitate the analysis: a nonlinear regression approach to more effectively deal with nonlinear patterns in the inﬂuent and efﬂuent data; and the regression of concentrations (ROC), which is an additional concentration-based metric. A statistical analysis of the results demonstrated that the differences between inﬂuent and efﬂuent streams for TSS are signiﬁcantly different in their mean and median, and the removal efﬁciency of the Ecosol Litter Basket was evaluated to be 57–65% for TSS with the inﬂuent event mean concentration (EMC) up to 142 mg/L.


Introduction
As evidenced by the recent release of Australian guidelines for the testing of stormwater filtration devices [1], increasing attention is focused on the protection of riverine ecosystems through the regulation of water quality requirements for discharge from stormwater systems. This study presents the field test results and filtration performance analysis of the Ecosol Litter Basket, in accordance with [1]. The main purpose of this research is to evaluate the performance of the Ecosol Litter Basket with respect to key pollutants. The field test detailed in this study was undertaken on a litter basket prototype installed in an urban catchment in Australia [2]. During the testing period from May 2017 to March 2019, a total of 29 rainfall events were recorded, of which between 13 to 16 events were evaluated as qualifying (for each pollutant type) for the purposes of characterizing the removal efficiency according to the methodology in [1]. The monitoring, verification of the testing methods adopted, and site evaluation were undertaken by The University of Adelaide.
The Ecosol Litter Basket is a primary stormwater filtration device, targeting gross pollutants and coarse to fine particulate matter. These devices are typically fitted to new and existing side entry pits but are customized to fit any stormwater inlet pit. All of the influent (stormwater runoff) entering the inlet flows through the litter basket mesh liner, where gross pollutants are captured and retained in the basket.
As part of a holistic approach to stormwater management, there exists a range of water-sensitive urban design strategies that use natural and engineered infiltration and The site is a part of the Redland City Council Depot, where the catchment area of the selected site is almost 423 m 2 and is predominately road and pavement, with 90% being impervious. Preliminary sampling was taken for water quality testing to characterize the trial site to avoid a site where pollutant concentrations were likely to fall outside the limits allowed for qualifying events, or where pollutant concentrations fall below laboratory limits of detection (LOD) [1].
The Ecosol Litter Basket captures pollutants conveyed by rainfall runoff at drainage entry points (at-source) and consists of a stainless-steel frame and removable filtration basket with a 200 μm mesh. It was designed as a primary treatment measure. For the testing unit installed at the test site, the treatable flow rate was estimated to be 22 L/s [2]. Figure 2 shows the installed device. The site is a part of the Redland City Council Depot, where the catchment area of the selected site is almost 423 m 2 and is predominately road and pavement, with 90% being impervious. Preliminary sampling was taken for water quality testing to characterize the trial site to avoid a site where pollutant concentrations were likely to fall outside the limits allowed for qualifying events, or where pollutant concentrations fall below laboratory limits of detection (LOD) [1].
The Ecosol Litter Basket captures pollutants conveyed by rainfall runoff at drainage entry points (at-source) and consists of a stainless-steel frame and removable filtration basket with a 200 µm mesh. It was designed as a primary treatment measure. For the testing unit installed at the test site, the treatable flow rate was estimated to be 22 L/s [2]. Figure 2 shows the installed device. The site is a part of the Redland City Council Depot, where the catchment area of the selected site is almost 423 m 2 and is predominately road and pavement, with 90% being impervious. Preliminary sampling was taken for water quality testing to characterize the trial site to avoid a site where pollutant concentrations were likely to fall outside the limits allowed for qualifying events, or where pollutant concentrations fall below laboratory limits of detection (LOD) [1].
The Ecosol Litter Basket captures pollutants conveyed by rainfall runoff at drainage entry points (at-source) and consists of a stainless-steel frame and removable filtration basket with a 200 μm mesh. It was designed as a primary treatment measure. For the testing unit installed at the test site, the treatable flow rate was estimated to be 22 L/s [2]. Figure 2 shows the installed device.

Instrumentation and Sampling Methodology
Sampling was undertaken by two Isco 3700 Full-size autosamplers (24 × 250 mL) for the influent and effluent streams during each rainfall. The pump rate of the Isco 3700 autosampler is 3000 mL/min with a line transport velocity of 0.70 m/s through a medicalgrade sample tube (3/8" inner diameter). For continuous depth measurement, pressure sensing LMP 307 stainless steel probes were used. A rain gauge (Model TB3) which was compatible with the requirements of [1] (sampling at intervals of 5 min and increments of 0.2 < 0.25 mm), was used for measuring rainfall. The data logger Model CR1000 was used for collecting the signals from the sensing instruments. The installed instrumentation is shown in Figure 3.

Instrumentation and Sampling Methodology
Sampling was undertaken by two Isco 3700 Full-size autosamplers (24 × 250 mL) for the influent and effluent streams during each rainfall. The pump rate of the Isco 3700 autosampler is 3000 mL/min with a line transport velocity of 0.70 m/s through a medicalgrade sample tube (3/8" inner diameter). For continuous depth measurement, pressure sensing LMP 307 stainless steel probes were used. A rain gauge (Model TB3) which was compatible with the requirements of [1] (sampling at intervals of 5 min and increments of 0.2 < 0.25 mm), was used for measuring rainfall. The data logger Model CR1000 was used for collecting the signals from the sensing instruments. The installed instrumentation is shown in Figure 3.

Instrumentation and Sampling Methodology
Sampling was undertaken by two Isco 3700 Full-size autosamplers (24 × 250 mL) for the influent and effluent streams during each rainfall. The pump rate of the Isco 3700 autosampler is 3000 mL/min with a line transport velocity of 0.70 m/s through a medicalgrade sample tube (3/8" inner diameter). For continuous depth measurement, pressure sensing LMP 307 stainless steel probes were used. A rain gauge (Model TB3) which was compatible with the requirements of [1] (sampling at intervals of 5 min and increments of 0.2 < 0.25 mm), was used for measuring rainfall. The data logger Model CR1000 was used for collecting the signals from the sensing instruments. The installed instrumentation is shown in Figure 3.    The adopted sampling methodology was selected based on the methodology discussed in [18]. Grab samples (a maximum of four samples at 2 min intervals at the start of each event) were used to evaluate event mean concentration (EMC) for total petroleum hydrocarbons (TPH) and total recoverable hydrocarbons (TRH). Following this, for composite samples, the volumetric intervals between the first 8 aliquots were selected to be 750 L. After collecting 8 aliquots, the volumetric sampling interval increased to 3000 L. Composite samples were used to evaluate event mean concentration (EMC) for TSS, TP, TN, and heavy metals. The locations of the sampling for influent and effluent streams are shown in Figure 4. The adopted sampling methodology was selected based on the methodology discussed in [18]. Grab samples (a maximum of four samples at 2 min intervals at the start of each event) were used to evaluate event mean concentration (EMC) for total petroleum hydrocarbons (TPH) and total recoverable hydrocarbons (TRH). Following this, for composite samples, the volumetric intervals between the first 8 aliquots were selected to be 750 L. After collecting 8 aliquots, the volumetric sampling interval increased to 3000 L. Composite samples were used to evaluate event mean concentration (EMC) for TSS, TP, TN, and heavy metals. The locations of the sampling for influent and effluent streams are shown in Figure 4.

Overview of Metrics
The removal efficiency metrics (suggested by [1]) that are generally used for evaluating the performance of a stormwater filtration device along with regression metrics are comprehensively discussed in [18] and are summarized in Table 1. The results of these two metric categories (efficiency and regression metrics) are presented and will be compared in Section 3.

Overview of Metrics
The removal efficiency metrics (suggested by [1]) that are generally used for evaluating the performance of a stormwater filtration device along with regression metrics are comprehensively discussed in [18] and are summarized in Table 1. The results of these two metric categories (efficiency and regression metrics) are presented and will be compared in Section 3.

Metrics Equation Equation
No.

Efficiency Metrics
Concentration Removal Effi- Efficiency Ratio (ER) Summation of Loads (SOL) Efficiency of Individual Storm Loads (ISL) Regression Metrics Regression of Loads (ROL) Regression of Concentrations With reference to Table 1: and are the influent and effluent event mean concentrations respectively; m is the number of storm events; is the total volume of flow for storm event j; Efficiency Ratio (ER)

Efficiency of Individual Storm Loads (ISL)
Individual Average Regression Metrics Regression of Loads (ROL) Regression of Concentrations (ROC) With reference to Table 1: EMC in and EMC out are the influent and effluent event mean concentrations respectively; m is the number of storm events; V j is the total volume of flow for storm event j; f (x) in Equation (8) is the nonlinear regression curve of the inlet loads with respect to outlet loads; f (x) in Equation (9) is the nonlinear regression curve of event mean concentration influent with respect to event mean concentration effluent; and x is the centroid of the area under the fitting curve.
To evaluate the fitting curve for both of the regression metrics, both the squared correlation R 2 and the root mean squared error (RMSE) are calculated, as below: where y i and y i are the measured and predicted outputs respectively (loading for ROL and event mean concentration for ROC), and µ is the average of the measured output.

Confidence Level
In accordance with [1], to achieve statistical confidence for the performance of the stormwater device, event mean concentration with relatively consistent concentrations and removal efficiencies may be required. In cases where the concentrations and removal efficiencies are more highly varied, additional samples are needed to account for the variability. The goal of testing is to satisfy the 90% statistical significance where practical to do so.

Results
The two key criteria for accepting a rainfall event as a qualifying event are based on the hydraulic properties of the run-off (i.e., duration, and the interval between events) and the input pollutant concentration values (i.e., concentrations within an acceptable band) [1]. These are comprehensively discussed in [18].
Between May 2017 and March 2019, a total number of 29 rainfall events occurred. Table A1 (in Appendix A) shows the recorded details of each event during the field test. It can be seen that a wide range of storm behaviors, and consequent flow rates were covered by the events. The total number of events for qualifying for each pollutant type is as follows: 15 for TSS, TN, and THM; 16 for TP; and 13 for TPH and TRH. Based on the sampling methodology, the event mean concentrations (EMCs) were determined by an independent NATA accredited laboratory analysis for both grab samples and the flowweighted composites of aliquots. Tables 2 and A2, Tables A3-A6 show the ECM results for all pollutant types for both the influent and effluent streams. In accordance with [1], effluent event mean concentrations less than the limit of detection (LOD) (reported by the laboratory) were set at half of the LOD. As observed in Table 2, the qualifying events cover a broad range of flow conditions from 1.84 L/s (total storm volume of approximately 5 kL) up to 32 L/s (total storm volume of above 17 kL), and influent TSS concentrations (from 6 mg/L up to 142 mg/L).

Statistical Analysis
A statistical analysis of the event mean concentration dataset is required for evaluating the performance of the stormwater filtration device. Within this section, firstly, data visualizations are presented, followed by an analysis of the influent and effluent distributions to determine the significance of the device on filtering the pollutant constituents.

Data Summary Plots
To provide an overview of the qualifying event data, Figure 5 provides interleaved bar plots of a direct comparison between the relative magnitude of the ECMs from the influent and the effluent streams. With regards to TSS, it is relatively clear that the effluent stream has reduced TSS concentration in comparison to the influent stream, particularly events 3, 4 and 7. Event 11 yielded anomalous results with the effluent marginally higher in TSS than the influent, which is attributed to the very low influent concentration and potential measurement errors. The results are less conclusive for the other pollutant types, with generally high relative effluent concentrations (as for TP, TPH and TRH), and even many cases where a higher effluent concentration was regularly recorded (as for TN and THM).

Statistical Analysis
A statistical analysis of the event mean concentration dataset is required for evaluating the performance of the stormwater filtration device. Within this section, firstly, data visualizations are presented, followed by an analysis of the influent and effluent distributions to determine the significance of the device on filtering the pollutant constituents.

Data Summary Plots
To provide an overview of the qualifying event data, Figure 5 provides interleaved bar plots of a direct comparison between the relative magnitude of the ECMs from the influent and the effluent streams. With regards to TSS, it is relatively clear that the effluent stream has reduced TSS concentration in comparison to the influent stream, particularly events 3, 4 and 7. Event 11 yielded anomalous results with the effluent marginally higher in TSS than the influent, which is attributed to the very low influent concentration and potential measurement errors. The results are less conclusive for the other pollutant types, with generally high relative effluent concentrations (as for TP, TPH and TRH), and even many cases where a higher effluent concentration was regularly recorded (as for TN and THM).  To reinforce the message from Figures 5 and 6 shows the box and whisker plots for influent and effluent ECMs of all pollutants (notches indicate the 95% confidence interval). As observed in this figure, all data are highly skewed towards the lower concentrations. Overlapping confidence intervals between the influent and effluent box plots indicate that  To reinforce the message from Figures 5 and 6 shows the box and whisker plots for influent and effluent ECMs of all pollutants (notches indicate the 95% confidence interval). As observed in this figure, all data are highly skewed towards the lower concentrations. Overlapping confidence intervals between the influent and effluent box plots indicate that the central tendency of the data may not be significantly different [18]. A significant difference in the central tendency of the influent and effluent streams (with the effluent being the lesser of the two) is a key indicator that the filtration device is effective in reducing pollutant concentrations. From the box plots, it is observed that only the TSS medians are clearly different, indicating the effectiveness of the device in filtering TSS. This is discussed in more detail in the following section. It is also noted that the variability in the data is quite high, and that the influent data typically have much higher variability than the effluent data, except for the case of TP.

Influent and Effluent Distributions, and Significance
In order to assess the significance in the difference between the influent and effluent streams (and the implied effectiveness of the device), first, the influent and effluent distributions need to be determined, followed by the use of an appropriate statistical significance test. Table 3 shows the descriptive statistics and distributional analysis results of the influent and effluent ECM data. The high coefficient of variation shows the high variability in the measurements, and it is observed that all sampled streams and ECMs were found to be lognormally distributed (Table A7 in Appendix B).
To better visualize the distributions, a comparison of the ECMs for the influent and effluent streams' cumulative probability plots for all paired datasets is shown in Figure 7. A visual consideration of Figure 7 indicates that only the TSS the influent and effluent distributions are clearly distinct from one another, which indicates that a meaningful reduction of TSS does occur from the influent to the effluent stream. For all other constituents, the difference between the influent and effluent streams is far less pronounced, as the regression lines for the influent and effluent streams are either very close, or even cross over one another.
With regards to the significance testing of the influent and effluent EMCs, as the data are lognormally distributed, the validity of the paired t-Test holds and is suitable for use (Table 4).
From the t-Test results, it can be seen that the influent and effluent EMCs for TSS, TP, TPH, and TRH are significantly different in their means with a 95% confidence. Further, the difference in the TSS ECMs was observed to be extremely significant with a p-value of 0.0002. This means that the Ecsosol Litter Basket is observed to remove fractions of these pollutant types. The results for TN and THM indicate no significant changes between the influent and effluent EMCs.
From the results of the nonparametric test, it is observed that the influent and effluent stream EMCs for constituents TSS, TP, TPH, and TRH are significantly different in their medians. The TSS differences in EMC from influent to effluent are considered extremely significant (with a p-value of 0.0002), where whereas the differences for TP, TPH and TRH are just classed as significant (with p-values on the order of 10 −2 ). The results for TN and THM show no significant changes between the influent and effluent EMCs (with p-values in excess of 0.5). These results confirm the observations from the data visualizations in Figures 5-7. The interpretation of this result is that only the performance metric results for TSS, TP, TPH and TRH in the following section are considered statistically significant. For completeness, nonparametric tests (Sign Test and Wilcoxon-Mann-Whitney Rank-Sign Test) are used for comparison, and were found to support the conclusions of the t-Test.

Performance Metrics Evaluation
As discussed in Section 2.3, for evaluating the performance efficiency of a stormwater filtration device, a range of pollutant removal metrics have been utilized in monitoring studies. In this section, the performance metrics for the Ecosol Litter Basket filtration of each pollutant type will be presented. The results for TN and THM were included for the sake of completeness (despite not being statistically significant).

Efficiency Metrics
Statistical parameters for the performance metrics, such as mean, median, standard deviation, coefficient of variation (CV), and confidence level are shown in Table 5. It can be seen that the mean value for TN and THM is highly negative, which can explain the results of the significance testing as discussed in Section 3.1.2, namely, the results for these pollutant types is highly variable, leading to a lack of significance in the results. The CV for all significant pollutant types is also very high, being around 50% for TSS, TPH and TRH, up to almost 200% for TP. This degree of variability in the results is typical for field studies of stormwater filtration devices. From Table 5, it is also noted that the confidence level for all pollutants, except TSS, is less than 90%. This means that despite all TSS, TP, TPH and TRH being evaluated as having significantly different influent and effluent ECMs, it is only TSS that provides high confidence in the performance metric values (classed in [1] as being greater than 90%).

Regression Metrics
Regression metrics include the regression of loads (ROL) and the regression of concentrations (ROC), as shown in Table 1 and comprehensively discussed in [18]. The best regression curves, along with their functions ( f (x)) for all pollutants loads, are shown in Figure 8. It is observed that the regression curve is linear only for TN, whereas for the other pollutant loads the regression curves are not linear, meaning that the influent-effluent relationship for these pollutant types is nonlinear. Table 6 shows a summary of the fitting curves along with the values of RMSE, R 2 , and ROL. The ROL efficiency for each pollutant was calculated based on Equation (8) at x = x, where x is the centroid of the area under the fitting curve. The low value of ROL for TN and THM is notable which is 14% and −12% respectively. It is also noted that the maximum ROL (65%) was achieved for TSS.
The best regression fitting curves for all event mean concentrations (EMCs) along with their functions (f (x) are shown in Figure 9. It can be seen that the regression curves for all pollutants are not linear. Table 7 shows a summary of fitting curves along with the values of RMSE, R 2 , and the ROC efficiencies. Note that ROC efficiency for each event mean concentration was calculated based on Equation (9) at where is the centroid of the area under the fitting curve. The low values of R squared and ROC for TN and THM are notable. As shown in Table 7, it can also be seen that Ecosol Litter Basket has the maximum ROC (65%) for TSS.    Note that regression efficiency results (ROL and ROC) will be valid if the loads and event mean concentrations through the device for a certain event are within the ranges shown in Tables 6 and 7.  Note that regression efficiency results (ROL and ROC) will be valid if the loads and event mean concentrations through the device for a certain event are within the ranges shown in Tables 6 and 7.

Concentration Removal Efficiency (CRE) Versus Influent Event Mean Concentration (EMC in )
Concentration removal efficiency (CRE) as a function of the influent event mean concentration (EMC in ) for all pollutant types are shown in Figure 10. Despite the large variability in the data, the observed trends are outlined in the following. It can be seen that for TSS, the CRE rises and reaches its median value (66.7%) at an influent concentration of 64 mg/L, for TP the CRE also rises and reaches the median CRE (28.1%) at an influent concentration of 0.07 mg/L, and for both TPH and TRH, the CRE falls quickly and reaches the median CRE (58.2% for TPH and 51.9% for TRH) at an influent concentration of 0.424 mg/L for TPH and 0.398 mg/L for TRH. For both TN and THM, concentration removal efficiency has a negative area and reaches the median CRE (21.4% for TN and 4.8% for THM) at an influent concentration of 2.9 mg/L for TN and 0.191 mg/L for THM. It is observed that these results for TN and THM are consistent with the other metrics results discussed previously.

Summary of the Ecosol Litter Basket Performance
Based on the formulas presented in Table 1, the ER and SOL for all qualifying events and for each pollutant calculated are shown in Table 8 and Figure 11. It is noted that the highest values of ER and SOL were achieved for TSS (62% for ER and 57% for SOL), whereas the lowest values of ER and SOL were achieved for TN (16% for ER and 12% for SOL) and THM (18% for ER and −11% for SOL). From the results, as highlighted in Figure 11, it is observed that there is a consistent agreement between the results of concentration and load efficiencies for TSS, TP, TN, TPH, and TRH. It is also observed that the required confidence level of 90% [1] was only achieved for the pollutant type TSS. It is important to note that the results for TN and THM are included for completeness, as evaluated in Section 3.1 the differences between the EMCs for the influent and effluent streams for TN and THM were not found to be statistically significant. By way of comparison, in [11] field testing results for a stormwater device consisting of a similar treatment train with a 200-micron mesh pit basket were presented. The site was predominantly roof area (56% coverage). Based on an analysis of nine qualifying events, the reported efficiency ratios were 32% for TSS, 37% for TP, and 38% for TN for the pit basket. Reference [10] completed a combined laboratory and field test study of the Ecosol Litter Basket. The mean value of collection efficiencies was evaluated to be 29% for TSS, 40% for TP, 11% for TN, 6% for THM, and 20% for hydrocarbons [10]. In [12], field testing of a gross pollutant trap (GPT) was undertaken over a period of two years at a commercial site located in Queensland, Australia. Approximately 85% of the total catchment area was impervious. The efficiency ratio (ER) calculated for the GPT was found to be 49% for TSS, 27% for TN, and 41% for TP. By way of comparison, in [11] field testing results for a stormwater device consisting of a similar treatment train with a 200-micron mesh pit basket were presented. The site was predominantly roof area (56% coverage). Based on an analysis of nine qualifying events, the reported efficiency ratios were 32% for TSS, 37% for TP, and 38% for TN for the pit basket. Reference [10] completed a combined laboratory and field test study of the Ecosol Litter Basket. The mean value of collection efficiencies was evaluated to be 29% for TSS, 40% for TP, 11% for TN, 6% for THM, and 20% for hydrocarbons [10]. In [12], field testing of a gross pollutant trap (GPT) was undertaken over a period of two years at a commercial site located in Queensland, Australia. Approximately 85% of the total catchment area was impervious. The efficiency ratio (ER) calculated for the GPT was found to be 49% for TSS, 27% for TN, and 41% for TP.

Conclusions
Within this paper, the results from a field-based performance analysis of the Ecosol Litter Basket are presented, where the results are based on the analysis of over 29 rainfall events, collected over a two-year period, within an urban catchment in Australia. Between 13 to 16 events were evaluated as qualifying for the purposes of characterizing the removal efficiency [1].
A variety of pollutant removal calculation methods, including concentration and load metrics, were utilized in this study to evaluate the efficacy of the performance of the device. In regard to the regression of loads (ROL), since a nonlinear regression is often more suitable for influent and effluent load data, a new approach presented by [18] was used for this metric. Regression of concentrations (ROC), as a new concentration-based metric introduced in [18], was also used to facilitate characterization of removal efficiency. It is strongly suggested that both concentration-based metrics (ER and ROC) and loadbased metrics (SOL and ROL) be calculated for characterizing the performance of a stormwater filtration device.
The statistical significance analysis showed the difference in event mean concentrations for the influent and effluent streams to be statistically significant for TSS, TP, TPH, and TRH. For the range of metrics considered, the statistically significant removal efficiencies of the Ecosol Litter Basket are summarized as: between 57-65% for TSS with the influent EMC up to 142 mg/L; 33-49% for TP with the influent EMC up to 0.32 mg/L; 28-51% for TPH with the influent EMC up to 3.82 mg/L; and 28-53% for TRH with the influent EMC up to 3.95 mg/L. Given the high variability observed in the EMC data, only the TSS results are estimated with high confidence.

Conclusions
Within this paper, the results from a field-based performance analysis of the Ecosol Litter Basket are presented, where the results are based on the analysis of over 29 rainfall events, collected over a two-year period, within an urban catchment in Australia. Between 13 to 16 events were evaluated as qualifying for the purposes of characterizing the removal efficiency [1].
A variety of pollutant removal calculation methods, including concentration and load metrics, were utilized in this study to evaluate the efficacy of the performance of the device. In regard to the regression of loads (ROL), since a nonlinear regression is often more suitable for influent and effluent load data, a new approach presented by [18] was used for this metric. Regression of concentrations (ROC), as a new concentration-based metric introduced in [18], was also used to facilitate characterization of removal efficiency. It is strongly suggested that both concentration-based metrics (ER and ROC) and load-based metrics (SOL and ROL) be calculated for characterizing the performance of a stormwater filtration device.
The statistical significance analysis showed the difference in event mean concentrations for the influent and effluent streams to be statistically significant for TSS, TP, TPH, and TRH. For the range of metrics considered, the statistically significant removal efficiencies of the Ecosol Litter Basket are summarized as: between 57-65% for TSS with the influent EMC up to 142 mg/L; 33-49% for TP with the influent EMC up to 0.32 mg/L; 28-51% for TPH with the influent EMC up to 3.82 mg/L; and 28-53% for TRH with the influent EMC up to 3.95 mg/L. Given the high variability observed in the EMC data, only the TSS results are estimated with high confidence.